RosettaCM input file for running on Stampede

What sort of errors are you getting? 

The biggest issue I see with the file is the extra “rosetta –thread threading5HT2c.txt –target 5HT2cfasta.txt –template 3PBL.pdb  –out 5HT2c.pdb –log 5HT2c.log” line at the end of it. I’m not sure what, if anything, you’re attempting to accomplish with that.

But other than that, it looks like you have the general idea – the SLURM file is simply a shell script with additional job control comments at the top of it. So you would write it almost exactly as you would write it if you weren’t submitting to a job control system. 

What sort of errors are you getting? 

The biggest issue I see with the file is the extra “rosetta –thread threading5HT2c.txt –target 5HT2cfasta.txt –template 3PBL.pdb  –out 5HT2c.pdb –log 5HT2c.log” line at the end of it. I’m not sure what, if anything, you’re attempting to accomplish with that.

But other than that, it looks like you have the general idea – the SLURM file is simply a shell script with additional job control comments at the top of it. So you would write it almost exactly as you would write it if you weren’t submitting to a job control system. 

What sort of errors are you getting? 

The biggest issue I see with the file is the extra “rosetta –thread threading5HT2c.txt –target 5HT2cfasta.txt –template 3PBL.pdb  –out 5HT2c.pdb –log 5HT2c.log” line at the end of it. I’m not sure what, if anything, you’re attempting to accomplish with that.

But other than that, it looks like you have the general idea – the SLURM file is simply a shell script with additional job control comments at the top of it. So you would write it almost exactly as you would write it if you weren’t submitting to a job control system. 

Okay, so those look like they’re issues with the TACC job submission system, rather than anything that’s Rosetta specific. I’d double check the TACC cluster documentation to make sure you’re doing things appropriately, and if you have any questions in that regards talk to the TACC cluster administrators. (The big issue to fix first off is the “Job submission is not allowed from this host.” error – again, consult the TACC documentation to figure out how you’re supposed to submit the runs.)

The one Rosetta-related thing that might also be happening is the “CANCELLED AT … DUE TO TIME LIMIT” this means that your runs are running longer than you’ve requested time for. My recommendation is to do a short (low -nstruct) pilot run, and look at the average CPU-time it takes to make a single structure (this should be printed to the Tracer). Then you can then multiply it out by the however many thousand structures you’re hoping to and calculate how long you’ll need on the number of processors you’ve allocated. Then you can set the requested runtime in the SLURM script accordingly. (I always pad the estimate to make sure my jobs don’t get killed.)

Keep in mind that most Rosetta runs are intended to be restartable, so if your job gets killed before it produces all the structure you need, you should be able to restart the command with exactly the same options and have it fill out the remainder. (It’s not exactly “pick up where you left off”, as when you restart you’ll typically pull a new random number seed, but it’s effectively the same.)

Okay, so those look like they’re issues with the TACC job submission system, rather than anything that’s Rosetta specific. I’d double check the TACC cluster documentation to make sure you’re doing things appropriately, and if you have any questions in that regards talk to the TACC cluster administrators. (The big issue to fix first off is the “Job submission is not allowed from this host.” error – again, consult the TACC documentation to figure out how you’re supposed to submit the runs.)

The one Rosetta-related thing that might also be happening is the “CANCELLED AT … DUE TO TIME LIMIT” this means that your runs are running longer than you’ve requested time for. My recommendation is to do a short (low -nstruct) pilot run, and look at the average CPU-time it takes to make a single structure (this should be printed to the Tracer). Then you can then multiply it out by the however many thousand structures you’re hoping to and calculate how long you’ll need on the number of processors you’ve allocated. Then you can set the requested runtime in the SLURM script accordingly. (I always pad the estimate to make sure my jobs don’t get killed.)

Keep in mind that most Rosetta runs are intended to be restartable, so if your job gets killed before it produces all the structure you need, you should be able to restart the command with exactly the same options and have it fill out the remainder. (It’s not exactly “pick up where you left off”, as when you restart you’ll typically pull a new random number seed, but it’s effectively the same.)

Okay, so those look like they’re issues with the TACC job submission system, rather than anything that’s Rosetta specific. I’d double check the TACC cluster documentation to make sure you’re doing things appropriately, and if you have any questions in that regards talk to the TACC cluster administrators. (The big issue to fix first off is the “Job submission is not allowed from this host.” error – again, consult the TACC documentation to figure out how you’re supposed to submit the runs.)

The one Rosetta-related thing that might also be happening is the “CANCELLED AT … DUE TO TIME LIMIT” this means that your runs are running longer than you’ve requested time for. My recommendation is to do a short (low -nstruct) pilot run, and look at the average CPU-time it takes to make a single structure (this should be printed to the Tracer). Then you can then multiply it out by the however many thousand structures you’re hoping to and calculate how long you’ll need on the number of processors you’ve allocated. Then you can set the requested runtime in the SLURM script accordingly. (I always pad the estimate to make sure my jobs don’t get killed.)

Keep in mind that most Rosetta runs are intended to be restartable, so if your job gets killed before it produces all the structure you need, you should be able to restart the command with exactly the same options and have it fill out the remainder. (It’s not exactly “pick up where you left off”, as when you restart you’ll typically pull a new random number seed, but it’s effectively the same.)